Re: impdp performance

I was informed offline this:

"IIRC indexes were not created in parallel but only one worker thread
creates them. this was logged as a bug/enhancement request. maybe be fixed
in 11g."

Is this true?

On Mon, Jun 22, 2009 at 3:31 PM, Niall Litchfield <
niall.litchfield@xxxxxxxxx> wrote:

> The alternative would presumably be to extract the ddl beforehand with
> dbms_metadata, you can then conditionally add parallel or not. Then run your
> script afterwards.
>
> On Mon, Jun 22, 2009 at 2:18 PM, Ram Raman <veeeraman@xxxxxxxxx> wrote:
>
>> Thanks Yechiel, That is what I did last time. I wanted to check if there
>> was something new.
>>
>> It was not a little edit though. It was not just search and replace. The
>> index definitions of some indexes had parallel already built in them, which
>> caused them to fail after I added my own parallel clause. I guess I will do
>> it that way.
>>
>>
>>
>> On Mon, Jun 22, 2009 at 12:32 AM, Yechiel Adar <adar666@xxxxxxxxxxxx>wrote:
>>
>>> You can use impdp with:
>>> sqlfile=index.sql,include=index
>>> to get the sql to create all the indexes.
>>>
>>> A little edit will make all the create index no logging and parallel ans
>>> analyzed.
>>> You get additional bonus as you can then split the index creation to 3-4
>>> parts and run them in parallel.
>>>
>>> Adar Yechiel
>>> Rechovot, Israel
>>>
>>>
>>>
>>> Ram Raman wrote:
>>>
>>>> Hi all,
>>>>  We have about 50,000 indexes in this schema. If I exclude indexes
>>>> during import and decide to create them later using Nologging and parallel
>>>> options, do I need to extract the index DDLs for each of the indexes and
>>>> modify the DDL for each of them or is there any quick way to set it for all
>>>> indexes.  v 10g.  I used a parallelism of 8.
>>>> Thanks.
>>>>
>>>>
>>> --
>>> http://www.freelists.org/webpage/oracle-l
>>>
>>>
>>>
>>
>
>
> --
> Niall Litchfield
> Oracle DBA
> http://www.orawin.info
>

Other related posts: